Session 1: Introduction and Visualisation Extended (Answers)
David Munoz Tord & Denis Mongin
October 10, 2022
Setup
Start up - Folder structure
(same as previous Rmd no need to do it twice)
Download datasets
- ExerciceBisphenol.csv
- ExercicePlasmaCholesterol.xlsx
And save them into your “data” folder.
Same .Rproject
Double check that you are still in the same Rproject (check on the top right of RStudio)
Load libraries
#load required packages
library(ggplot2) #to plot
library(readr) #to read files
library(here) #to find files
#if you want to make it reproducible (more advance)
# if(!require(pacman)) {install.packages(c("pacman"))}
#
# pacman::p_load(here,ggplot2, readr)
#these are just Rmarkdown visual options
knitr::opts_chunk$set(message=F, warning=F, comment=NA) Find your path
#here() #from the here packages is a neat magick trick that will find where you files are
#given that you are in a .Rproject, type help(here) for more infoRead the data
il18 <- read_csv(here("data", "freeIL18.csv")) #read the data
#you can also just read it manually
#il18 <- read_csv("C:/Users/david/Documents/cours_stat/freeIL18.csv")
il18# A tibble: 73 x 4
samples `Disease activity` `Free IL-18 (pg/ml)` `Total IL-18 (ng/ml)`
<dbl> <chr> <dbl> <dbl>
1 1 A 1.37 0.544
2 2 I 1.37 0.145
3 3 I 1.37 0.0893
4 4 A 1.37 1.50
5 5 A 1.37 0.167
6 6 I 1.37 0.0478
7 7 A 75.5 22.5
8 8 I 1.37 1.23
9 9 A 187 36.1
10 10 I 110 30.0
# ... with 63 more rows
# i Use `print(n = ...)` to see more rowsnames(il18) <- c("samples", "dis.act", "freeIL18", "TotalIL18") # assign new names to columnsVisualizations
Numerical
Density Plots
Show the density of a numeric variable
A density plot shows the distribution of a numeric variable. It takes only numeric variables as input and is very close from an histogram. It can be use in the same exact condition.
Tips:
Play with the bandwidth of your density function.
Don’t show the distribution of more than ~5 variables. Use Violin or Ridge line plot instead.
Avoid filling with color palettes.
Do a Basic density plot: x = freeIL18
#hint: ggplot(data, aes(...)) + geom_density()
#hint2: assign your plot to an object, e.g. density <- ggplot(data, aes(...)) + geom_density()
density <- ggplot(data = il18, aes(x = freeIL18)) +
geom_density( fill="green")
density
Density plot by groups (dis.act)
#hint use fill
ggplot() +
geom_density(data = il18,
aes(x = freeIL18, fill=dis.act)) 
Stacked density plot
#hint: same as before but use position="fill" in geom_density
ggplot() +
geom_density(data = il18,
aes(x = freeIL18, group=dis.act, fill=dis.act), position="fill")
Data Manipulation
Load libraries
#install.packages("tidyr") # data manipulation
#install.packages("readxl") # read excel files
#install.packages("writexl") # write excel files
#install.packages("scales") # work with different scales
#<- you just need to run these lines once
library(tidyr)
library(readxl)
library(writexl)
library(scales)Try to read a .csv file: ExerciceBisphenol.csv
#hint: use read_csv() and call it df_bisphenol
df_bisphenol <- read_csv(here("data", "ExerciceBisphenol.csv")) #read the data
#you can also just read it manually
#df_bisphenol <- read_csv("C:/Users/david/Documents/cours_stat/ExerciceBisphenol.csv")
df_bisphenol# A tibble: 3 x 3
group Cancer NoCancer
<chr> <dbl> <dbl>
1 Control 1 29
2 Moderate exposition 2 28
3 High exposition 10 20Try to read an excel file
#hint: use read_xlsx() and call it plasma
plasma <- read_xlsx(here("data", "ExercicePlasmaCholesterol.xlsx"))
#you can also just read it manually
#plasma <- read_xlsx("C:/Users/david/Documents/cours_stat/ExercicePlasmaCholesterol.xlsx")
plasma# A tibble: 20 x 3
Control `Treatment 1` `Treatment 2`
<dbl> <dbl> <dbl>
1 164. 191. 164.
2 173. 171. 153.
3 178. 184. 154.
4 170. 172. 152.
5 183. 176. 158.
6 172. 168 156.
7 188. 182. 157
8 178. 160. 160.
9 186. 173. 160.
10 176. 170. 148.
11 171. 177. 153.
12 177. 162. 158.
13 171. 176. 158.
14 182. 177. 156.
15 179. 165. 157.
16 187. 160. 154.
17 166 165. 147.
18 178. 173. 159.
19 167. 169. NA
20 172. 172. NA Look at the data
Output the summary of the Bisphenol data
#hint use summary() and or str()
summary(df_bisphenol) group Cancer NoCancer
Length:3 Min. : 1.000 Min. :20.00
Class :character 1st Qu.: 1.500 1st Qu.:24.00
Mode :character Median : 2.000 Median :28.00
Mean : 4.333 Mean :25.67
3rd Qu.: 6.000 3rd Qu.:28.50
Max. :10.000 Max. :29.00 Same for the Plasma data
#hint use summary() and or str()
summary(plasma) Control Treatment 1 Treatment 2
Min. :164.4 Min. :159.5 Min. :147.4
1st Qu.:170.8 1st Qu.:167.2 1st Qu.:153.4
Median :176.3 Median :171.9 Median :156.8
Mean :175.9 Mean :172.0 Mean :155.7
3rd Qu.:179.9 3rd Qu.:176.2 3rd Qu.:157.7
Max. :187.9 Max. :190.8 Max. :163.6
NA's :2 Let’s look deeper into plasma
We will start with plasma. We have names with spaces, let’s rename them !
#rename the columns of plasma data to "control","treatment_1","treatment_2"
#hint: use names() and a vector of names i.e. c("..", "..")
names(plasma) <- c("control","treatment_1","treatment_2")
plasma# A tibble: 20 x 3
control treatment_1 treatment_2
<dbl> <dbl> <dbl>
1 164. 191. 164.
2 173. 171. 153.
3 178. 184. 154.
4 170. 172. 152.
5 183. 176. 158.
6 172. 168 156.
7 188. 182. 157
8 178. 160. 160.
9 186. 173. 160.
10 176. 170. 148.
11 171. 177. 153.
12 177. 162. 158.
13 171. 176. 158.
14 182. 177. 156.
15 179. 165. 157.
16 187. 160. 154.
17 166 165. 147.
18 178. 173. 159.
19 167. 169. NA
20 172. 172. NA Wide to long format
Wide, or unstacked data is presented with each different data variable in a separate column.
Long, or narrow data is presented with one column containing all the values and another column listing the context of the values.
Many functions in R expect data to be in a long format rather than a wide format.
Here the data is in a wide format, let’s put it in long format so we can use all the power of mist R functions.
For that we use the package tidyr
plasma_long <- pivot_longer(data = plasma,c("control","treatment_1","treatment_2"))
plasma_long# A tibble: 60 x 2
name value
<chr> <dbl>
1 control 164.
2 treatment_1 191.
3 treatment_2 164.
4 control 173.
5 treatment_1 171.
6 treatment_2 153.
7 control 178.
8 treatment_1 184.
9 treatment_2 154.
10 control 170.
# ... with 50 more rows
# i Use `print(n = ...)` to see more rowsReflect: Do you see the differences ?
Now do a boxplot : x = name, y = value and add point
#hint: use geom_boxplot + geom_jitter
#hint2: you might want to rename the column ?
#bonus: try to do the same plot with plasma (the one in wide format)
ggplot(plasma_long,
aes(x = name,y= value))+
geom_boxplot()+
geom_jitter()
Let’s look deeper into bisphenol
Contingency tables of some experiment, to check is groups are of the same length.
df_bisphenol$tot <- df_bisphenol$Cancer + df_bisphenol$NoCancer
df_bisphenol# A tibble: 3 x 4
group Cancer NoCancer tot
<chr> <dbl> <dbl> <dbl>
1 Control 1 29 30
2 Moderate exposition 2 28 30
3 High exposition 10 20 30Let’s try a barplot with this wide format data:
First just plot the “cancer”
NB: we use geom_col
wide_plot <- ggplot(df_bisphenol)+
geom_col(aes(group,Cancer,fill = "cancer"))
wide_plot
Then add the “no cancer”
wide_plot+
geom_col(aes(group,NoCancer,fill = "no cancer"))
#### Problem!
The second bar is over the first one! Need to be in long format too! Can you do it ?
#hint: use pivot_longer
df_bisphenol_long <- df_bisphenol %>%
pivot_longer(c("Cancer","NoCancer"))
# names(df_bisphenol_long) <- c("group","output","Nindividual")
ggplot(df_bisphenol_long,
aes(group,value,fill = name))+
geom_col()+
coord_flip()
Change from “stacked” plot to “dodged” plot
#hint: do the same thing but add position = "dodge" to geom_col()
ggplot(df_bisphenol_long,
aes(group,value,fill = name))+
geom_col(position = "dodge")+
coord_flip()
Add percents
#hint do the same thing as step one but add scale_y_continuous(labels = percent_format(accuracy = 1)) at the end
ggplot(df_bisphenol_long,
aes(group,value,fill = name))+
geom_col(position = "fill")+
coord_flip()+
scale_y_continuous(labels = percent_format(accuracy = 1))
More Visualisation Stuff
Use themes
Default
violin <- ggplot(data = il18, aes(x = dis.act, y = freeIL18, fill=dis.act )) +
geom_violin() +
xlab("Disease activity")
violin + theme_gray() # the default
Black and White
violin + theme_bw() # the default
Classic
violin + theme_classic() # the default
Interactive plots
The package plotly helps creating awesome interactive data vizualisation and maks it super easy to do !
First load it
#install.packages(plotly) # install it if you don't have it
library(plotly)Lets redo the last scatter plot but this time with a twist !
plot_object <- ggplot() +
geom_point(data = il18,
aes(x = freeIL18, y = TotalIL18, color=dis.act)) +
geom_point()
ggplotly(plot_object)